[jl6]

When the project started, CPU frequency scaling wasn't a thing, so CPUs would run at full speed (and full power draw) 100% of the time. If you weren't making maximal use of the CPU, any remaining capacity would go to waste. Distributed computing projects could make use of that remaining capacity.

Today, CPUs are built with power efficiency in mind, and will attempt to scale down rapidly when not fully in use. Thus there is no longer such a thing as "spare CPU time". Any time spent on distributed computing projects is paid for in electricity costs. Some choose to continue anyway, but many have been disincentivized.

[deadbeeves]

I don't think that's true. Variable frequency certainly helps efficiency, but like the other commenter said, HLT did exist. The CPU would use less power when told by the OS to do nothing for a short while.

[organsnyder]

For a while I had a Home Assistant automation that would spin up Prime95 on a machine in my homelab when the closet it was in (in an unheated garage) got too cold. The closet also has the water meter, so it has to be kept above freezing. There's also a resistive heater, but I figured I'd rather get a bit of productive use out of those watts.

Then I realized that the computers heated the closet plenty without artificially pegging CPUs, so I didn't bother reimplementing it when I did a migration.

[tedunangst]

The hlt instruction has been around for a while.

[post-it]

I don't think that saved any power on a CPU without frequency scaling.

[spookthesunset]

Hmmm.... there was some windows program called "waterfall" that kept your CPU cool when it was idle. Very useful for overclockers.

[psd1]

Was it snake oil like memory compressors?

Are some instructions cooler than others?